BREAKING
Local Deep Research goes fully offline
An agentic search-and-verify loop
1Search
2Summarize
3Verify
4Cited report
Local-first, cloud optional
Local runtimes
Ollama, llama.cpp
Encrypted local library
No cloud API required
Optional cloud
OpenAI, Anthropic, Google
Cost & token tracking
Docker & Unraid support
Benchmark scores on local models
SimpleQA large95.7
SimpleQA small91.2
xbench-Deep77
Small samples, hallucination risk
Private research, no cloud needed
AI NEWS BLITZ
An open-source tool now generates cited research reports entirely on local LLMs.